Prediction of Water Inrush in Long-Lasting Shutdown Karst Tunnels Based on the HGWO-SVR Model

Dunwen Liu,Yu Tang,Qian Xu,Yinghua Jian

doi:10.1109/access.2020.3047626

Abstract

It is very important to predict the long-term shutdown karst tunnel water inrush for preventing tunnel construction accidents. However, it is urgent to study a new prediction model to solve the problems of insufficient sample size and low prediction accuracy for long-term shutdown karst tunnel water inrush prediction. In this study, the water inrush and atmospheric rainfall in a tunnel project in China were monitored for over five months. By adopting hybrid grey wolf optimization (HGWO) algorithm and support vector regression (SVR) method, the HGWO-SVR tunnel water inrush prediction model was proposed. The atmospheric rainfall of the day and yesterday and yesterday’s water inrush were considered in the HGWO-SVR model, and the model was used to predict the tunnel water inrush. The results show that the predicted water inrush value is basically consistent with the measured value. After the parameters of SVR model are optimized by HGWO algorithm, the HGWO-SVR prediction model has the advantages of high precision and less sample demand. The model is more suitable for the prediction of long-term shutdown tunnel water inrush with less measured sample. Thus, the proposed prediction model can effectively be used as a new approach for tunnel water inrush in some similar projects.

Highlights

Due to the development of construction for highway tunnels in China gradually advanced to the southwest, a large number of water inrush problems are becoming increasingly serious
In this study, an intelligent water inrush prediction algorithm based on hybrid grey wolf optimization (HGWO)-support vector regression (SVR) is proposed for long-lasting shutdown tunnel
According to the data collected from field measurements, the improved HGWO-SVR intelligent prediction algorithm was used to predict the water consumption in a long-lasting shutdown tunnel

Summary

INTRODUCTION

Due to the development of construction for highway tunnels in China gradually advanced to the southwest, a large number of water inrush problems are becoming increasingly serious. D. Liu et al.: Prediction of Water Inrush in Long-Lasting Shutdown Karst Tunnels Based on the HGWO-SVR Model analyzed the advantages and disadvantages of each prediction method and the applicable conditions [4]. A mixed gray wolf optimization support vector regression (HGWO-SVR) prediction model is proposed, choosing the three factors of yesterday’s rainfall, today’s rainfall and yesterday’s water inrush. To obtain the best combination of penalty factor C and kernel parameter g, the HGWO algorithm is employed to perform a fast global optimization search It can reduce the blindness of trial calculations and improve the accuracy of model prediction. D. Liu et al.: Prediction of Water Inrush in Long-Lasting Shutdown Karst Tunnels Based on the HGWO-SVR Model FIGURE 2. A clear linear relationship between precipitation and the amount of water inrush in the tunnel area can be seen in the figure

DATA PROCESSING AND ANALYSIS

CONCLUSION